Preliminary investigation into the feasibility of combining satellite and GPS data to identify pasture growth and grazing

Abstract

Regular estimation of pasture availability is a time-consuming on-farm task, but one that is vital for good grazing management. The ability to automate this task is, therefore, highly valuable. Combining satellite sensing of pasture mass with global positioning for herd location provides raw data that can potentially be used to automatically estimate pasture mass, pasture growth and pasture grazing events across a farm. The feasibility of automatically obtaining and processing this information was demonstrated on a Waikato dairy farm from 22 October 2018 to 21 February 2019 (123 days), with 13 global positioning collars recording the location of grazing mobs 16 times per hour on average, in a dairy herd of initially 380 animals. Satellite sensing of pasture cover over the same period was only possible on 16 days during this period, with November being particularly cloudy, resulting in fewer pasture cover estimates. A non-linear regression model was constructed with parameters representing initial pasture cover, average pasture growth rate through time, pasture growth differences between paddocks, pasture disappearance rate relative to the density of cow GPS samples, and an ungrazeable residual. A Bayesian approach was used to infer the model parameters from the satellite-measured pasture cover data. This allowed interpolation of pasture mass through the whole period with an RMSE of 225 kgDM/ha, as well as identifying growth rate differences between paddocks, which may provide a useful basis for improved pasture management. Rough estimates of cow average daily pasture disappearance were also made, which peaked at 20 kgDM/d in November, falling to 5 kgDM/d by February. This pilot study demonstrated the feasibility of combining satellite pasture cover data with herd location data from a small number of GPS collars to infer pasture growth rates in individual paddocks through time.